JP5147293B2 - Bill evaluation method and apparatus - Google Patents

Description

  The present invention relates to a method and apparatus for evaluating banknotes.

  It is well known to evaluate a banknote by measuring the optical properties of the banknote, and to process this measurement along with acceptance criteria to determine whether the banknote belongs to a given class, i.e. face value. Yes. Bills can be scanned and reflected or transmitted, or both reflected and transmitted, and can be used to measure optical properties. The properties of banknotes at different light wavelengths (some or all of which may be invisible) can be measured.

  The characteristics of device components such as light emitters or sensors can vary from device to device and from time to time. For this reason, the sensor of the apparatus cannot be relied upon for stable and predictable measurement.

  It is known to mitigate this problem by frequently calibrating the device. Various calibration techniques can be used. For example, in a reflective system, when no banknote is present, the reflective surface can be used to calibrate the banknote path from the light emitter and sensor so that calibration measurements can be made by illuminating the surface and detecting the amount of light reflected back to the sensor. It may be provided on the opposite side. This calibration measurement can be used to adjust the intensity of the light emitted from the light emitter and / or the gain applied to the signal from the sensor so that a predetermined measurement can be made.

  This technique is not readily applicable to systems where light transmission through bills is measured because the reference plane interferes with the light path. One solution involving a movable reference plane for this is disclosed in EP-A-0731737. Alternatively, the reference plane may take the form of a calibration sheet that is moved into the bill path when a calibration measurement is performed.

  EP-A-0679279 discloses a device for detecting counterfeit bills in which bills are sucked in manually over a light emitter and a sensor is housed in a unit having a glass window. In this device, the light intensity is monitored by detecting the amount of radiation reflected inward from the glass window. However, such devices are also not suitable for transmission systems.

  These calibration techniques also suffer from several additional disadvantages. For example, when using a calibration sheet, the calibration operation is inconvenient when frequent calibration is required, requires improper manual execution, or, in the case of automatic, complex sheet drive structures. Must be provided. Also, although calibration techniques rely on a reference surface with stable optical properties, this is not necessarily so. For example, the optical characteristics may change due to dirt such as dust.

  It is also known to reduce the component variation problem by performing normalization of sensor measurements. See, for example, EP-A-0560023. Each bill is scanned along a different track. Within each track, for each color being measured, measurements are made through that track using the same components. This measurement is normalized by taking the ratio of the measured value to the sum of the measured values for the same color of the entire track scanned along the bill (“spatial normalization”). Therefore, the influence of component variations is reduced.

However, such spatially normalized measurements are relatively insensitive to relative amounts of different colors and are therefore not suitable for accurate bill identification. Therefore, the measurements were normalized using another technique. According to this further technique, the measurement of different colors within a particular area was normalized by deriving the ratio of each measurement and the sum of all measurements for different colors in that area. This “spectral normalization” technique is useful for appraisal because it preserves color information. This technique also makes the measurement practically less sensitive to the brightness of each area of the banknote, and therefore less sensitive to the amount of dust on the banknote. Thus, new (undirty) and old (dirty) banknotes have a small measurement variance, which improves recognition capabilities. However, spectrally normalized measurement values are more responsive to component variations and the luminance information is reduced, so the measurement values are not effective in determining the face value.
EP-A-0731737 EP-A-0679279 EP-A-0560023 EP-A-1321904

  Thus, while current banknote evaluators address the issue of component variation, both normalization and calibration methods benefit from improvements, particularly but not exclusively, in transmission systems.

  Aspects of the invention are set out in the accompanying claims.

  The present invention provides an alternative solution to the above problem. One solution provides a method by which calibration measurements can be easily performed even in devices that rely on transmission methods. Other solutions include a normalization method that reduces the data loss of the normalization method while also compensating for component variations. In principle, each method would be useful without the use of other techniques (and this application is intended to cover such use), but as described below, these techniques can be combined. The use is particularly synergistic.

  According to the first aspect of the present invention, the banknote evaluation apparatus can measure the optical characteristics of banknotes using light that is irradiated by a light emitter and reflected (preferably diffusively) by a banknote to a sensor. A light emitter and a sensor are provided on one side of the bill passage. There is also an optical device on the opposite side of the banknote passage that makes it possible to measure the transmission properties of the banknote. The optical device may be a second sensor that receives light from a light emitter on the opposite side of the banknote path, or a second light emitter that transmits light through the banknote to a sensor on the other side of the banknote path. There may be. The optical device has a window overlying and disposed between the device and the bill path. Calibration measurements are made using light from a light emitter on the first side of the passage, which traverses the bill passage and is then reflected by the window to the sensor on the first side of the passage.

  On the second side of the passage there may be more than one optical device, for example an emitter and a sensor, to allow reflection and transmission measurements to be made on both sides of the bill passage. Each optical device has a window. The optical devices may share a common window.

  By using windows, calibration readings can be easily made despite the fact that the device is organized to make transmission measurements. This window helps to prevent dust and dirt that collects on components associated with the optical device, such as the optical device and / or lens. Dust may collect on the window itself, but this can be easily cleaned, especially if the window is flat.

  Using such a calibration technique, it is possible to compensate for component variations that would result in relative color level uncertainty. This is accomplished by considering calibration measurements that indicate the relationship between different color levels.

  Assuming this window is known and has a constant reflectivity, it will also allow compensation for component variations that affect the measurement of luminance levels. However, when using windows for calibration measurements, it may be difficult to ensure that such conditions are met, especially if the windows may collect dust.

  According to the second further aspect of the present invention, the banknote evaluation apparatus determines a range in which each measurement value is different from an average value of a plurality of measurement values of different colors at each of a plurality of different positions of the banknote. Organized to normalize the optical measurements of the bills. This normalized value may be a function of the ratio of the measured value to the average value. Thus, the measurements are normalized both spectrally and spatially. The apparatus preferably comprises a plurality of sensors, each scanning a banknote track, and each measurement is normalized using other measurements performed by the same sensor.

  Combining both spectral and spatial normalization reduces information loss while also providing a degree of component variation compensation. In particular, normalized measurements will retain almost all information related to relative color levels, which improves spatial normalization. In addition, the normalized measurement value is relatively insensitive to the entire luminance level over the entire region including the measurement value on which the normalization is based. Insensitivity to the luminance level (i) compensates for component variations that affect the measured luminance (and thus improves over spectral normalization), and (ii) differs Reduce measurement variations (to some extent) resulting from banknotes in state. Thus, a single normalization operation (resulting in a single set of measurements for processing) is compared to two separate spectral and spatial normalization methods of the prior art. And can make a profit. However, component variations that affect the relationship between color measurements may not be fully compensated. Furthermore, certain advantages of separate spectral and spatial normalization steps are mitigated by combining the two steps. It is therefore not clear that such a combination provides an overall advantage. However, it has now been recognized that significant advantages can be obtained, especially (but not exclusively) when combined spatial / spectral normalization is used with certain calibration methods.

  It will be particularly understood that it is particularly advantageous to combine the first and second further aspects of the invention. The calibration method of the first further aspect is inexpensive, easily implemented, and component variations that may not be fully processed by the normalization method of the second further aspect (ie, the relationship between color measurements). It is possible to compensate for the variation of the influencing components. On the other hand, measurements that are normalized based on the large number of colors distributed over a substantial area of the banknote are relatively insensitive to the overall brightness level, so this calibration method does not match the measured brightness level. If the influencing component variations are not compensated for, these component variations are instead processed by a normalization method. Thus, the described calibration and normalization process provides a stable and predictable measurement that preserves a large amount of information related to relative color levels.

  Thus, a preferred embodiment of the present invention uses light reflected by a light emitter and sensor on one side of the bill path from the window on the other side of the bill path and above the further optical device. A device where calibration measurements are made and normalizes at least some of the measurements made using the light emitters and sensors and relates to multiple wavelengths at multiple locations where the measurements extend over the bill A value representing a range different from the average of a large number of measured values is obtained.

  The invention also extends to a third further aspect, according to which a measurement of a particular wavelength at a particular position is taken at one or more positions (including said particular position) Normalized with respect to a measurement group that includes measurements at multiple wavelengths (which includes a plurality of wavelengths), and the normalized measurement value represents a relationship between the measurement value and a value that represents the variance of the measurement values within the group. The Preferably, the normalized measurement is a ratio of (i) the difference between the measurement and the average measurement of the measurement group and (ii) a value representing the variance of the measurement within the group. Can be obtained. The variance value may be a standard deviation of measured values. Such a technique provides the advantage of reducing the effect of measured value dispersion due to variations in ink density caused by wear or as a result of changes in printing conditions.

  This normalization method of the third further aspect of the invention may be used instead of the normalization method of the second further aspect of the invention. Alternatively, (a) perform a single normalization operation according to the second and third further aspects of the present invention, or (b) perform different normalization methods and apply individual acceptance criteria. Both methods may be used, either by deriving individual sets of measurements.

  A configuration embodying the present invention will now be described by way of example with reference to the accompanying drawings.

  FIG. 1 schematically shows an automatic transaction system (such as a vending machine) 3 provided with a bill evaluation apparatus 1 according to the present invention. The evaluation device has at least one receiving opening 11 and at least one drawer opening 12 for receiving and returning bills, and also has a measuring unit 13, a determination unit 14 having a data store 30, a control unit 15, a plurality of one-way storage devices 16 to 16i, and a plurality of two-way storage devices 17 to 17i. These units are connected by means of transport 20, 21, 22, 23, 24, 25 and a common routing element 18.

  When the banknote 2 is inserted into the receiving opening 11, the banknote is conveyed by the first transport means 20 to a measuring unit 13 including a measuring device necessary for checking acceptability and determining a face value. The measured value is then processed using the data stored in the data store 30 to determine whether the bill is acceptable, and if it is acceptable, the bill is reused. Sent to a decision unit 14 that determines whether the type is assigned. The control unit 15 is commanded to control the common routing element 18 of the transport system accordingly. Upon exiting the measurement unit 13, the unacceptable banknotes are returned directly to the drawer opening 12. Acceptable banknotes that are not reused are directed by the routing element 18 onto the transport means 23 and transported to one of several one-way stores 16-16i. Acceptable banknotes available for reuse are guided by the routing element 18 onto the vehicle 24 and carried to one of several bi-directional stores 17-17i for storage.

  The unit 15 can control the two-way storages 17-17 i to supply a desired type and number of banknotes 2 to the drawer opening 12 via the transport means 25.

  The banknote evaluation apparatus 1 corresponds to a prior art apparatus and operates as follows. Each banknote received at the receiving opening 11 is measured in the unit 13 using an optical test that includes determining the reflectivity and transmission of the banknote in different regions and different spectral regions. In order to derive a large number of measurements, the banknotes are preferably scanned in an area distributed substantially over at least one surface, preferably on both sides.

  The unit 14 then uses stored data from the storage 30 representing several different target classes, each corresponding to an individual genuine face value, and possibly other target classes corresponding to known counterfeit bills. Process those measurements. Many suitable processing techniques are known to those skilled in the art. As is well known, this test procedure generally involves using separate tests (using different data) to determine whether a received bank note belongs to each individual target class or face value.

  If the determining unit 14 determines with a certain level of certainty that the received banknote belongs to the real target face value, a suitable signal is sent to the control unit 15. Subsequently, the control unit 15 sends a signal in a control section (not shown) of the automatic transaction machine 3 via the bidirectional path 19. The transmitted signal represents a credit amount given to the user in exchange for the received bill.

  The automatic transaction system 3 preferably incorporates the display device 32 and is organized so that the display device 32 displays the credit amount given to the user.

  A real banknote is sent to a suitable one of the storages 16-16i or, if the banknote has a refillable and dispensable face value, to one of the two-way storages 17-17i. .

  After the transaction, for example, after the vending operation, the automatic transaction machine 3 can send a signal on the route 19 to cause the control unit 15 to refund the predetermined amount from the two-way storage units 17 to 17i.

  The measurement unit 13 is preferably geometrically organized as described in EP-A-1321904 (incorporated herein) for performing reflection and transmission measurements, but in connection with the details described below. EP-A-1312904 is different. The optical properties of the banknotes are measured using pairs of knitted modules, where the banknote path passes between each pair of modules. FIG. 2 shows an exemplary pair of modules 200 </ b> A and 200 </ b> B during the process of scanning banknote 2.

  In the illustrated embodiment, each module is along a line that is parallel to the width dimension of the bills passing between the modules and that traverses the transport direction indicated by arrow A (preferably perpendicular to the transport direction). It comprises three optical units 202 positioned side by side.

  As shown in FIG. 2, the optical unit 202 of each module faces the corresponding unit 202 of the opposite module. The optical unit of the module is organized to irradiate and receive light traveling in a plane extending between the two modules, which is inclined with respect to the plane of bill transport.

  Each optical unit includes a light emitter 220 disposed between two sensors 240. Each light emitter directs light to the area of the bill that passes between the modules, and the bill reflects the light diffusely to the adjacent sensor 240. Exemplary light rays emitted from the light emitter are shown at 260 and diffusely reflected light is shown at 280. Each light emitter 220 is also operable to transmit light through bills to a pair of sensors 240 in the corresponding unit 202 of the opposing module on the opposite side of the bill passage, for example as shown at 290. In this way, as each bill passes between modules while moving in the direction of arrow A, each sensor scans each line extending along the bill, while scanning a number of lines along the scanned line. Perform reflection and transmission measurements at the locations.

  Each light emitter includes a number of light emitting dies (not shown) that emit light of different colors. These are driven continuously. Thus, each sensor 240 can detect the reflection and transmission characteristics of each different wavelength at a number of locations along the line being scanned by the sensor, and the sensor 240 can be along a line that traverses it. Are arranged.

  This is a preferred embodiment, but in other cases each emitter may have a configuration that emits light extending over a broad spectrum, and each sensor detects light in a limited wavelength band. A number of individual receiving elements, each provided with a filter.

  FIG. 3 is a longitudinal cross-sectional view of one of the modules. Each module includes a substrate formed by a circuit board 320. The circuit board has a rear surface 324 and an opposite front surface 326, both of which carry electronic components (not shown in FIG. 3 for clarity). The components on the front side of the circuit board include light emitting components forming each light emitter and light receiving components forming each sensor. The plastic housing 328 is attached to the front surface of the circuit board, and the housing is formed with openings for exposing the light emitting and light receiving components, a portion of which is shown at 330. A collimated light emitter lens 332 is supported by the housing and resides on the light emitter element. Similarly, a collimating sensor lens 334 is supported by the housing and resides on the sensor component. The housing also supports an elongated window 336 at its front end, which is preferably made from a transparent plastic material such as polycarbonate, eg, polymethylmethacrylate, over the light emitting lens and sensor. Exists. The housing is formed with a separating wall, some shown at 338, which ensures that light from each of the light emitters cannot be reflected by the window to the adjacent sensor.

  The window 336 of each module faces the opposite module. The window is transparent, but some of the light incident on the window is reflected instead of being transmitted. Thus, if no banknotes are present between the modules, the light from each module's light emitter can reach the opposite module's window, and some of that light is re-directed (specularly) directly to the first module. Reflected. About 5% of this light is reflected by the air / window interface in front of the window as shown at 340, and about 5% of the remaining light is specularly reflected by the window / air interface behind the window as shown at 342. Re-reflected. A portion of the specularly reflected light reaches a sensor adjacent to the light emitter through which the light is transmitted. The specular light received by this sensor is measured during the calibration operation.

  Here, the operation of the measurement unit will be described.

  When the banknote 2 is inserted into the opening 11, the banknote is detected by a detector (not shown), operates the transport system to the control unit 15 to deliver the banknote to the measurement unit 13, and sends a certain signal. Send to measurement unit to start calibration operation. The calibration operation is performed before the banknote 2 has time to reach the measuring unit 13.

  The calibration operation involves each multicolor light emitter being driven to emit light of each wavelength continuously. For each wavelength, a reading is taken from each adjacent sensor. Assuming that the number of wavelengths is C and there are N sensors, C × N calibration readings are obtained.

  Using the calibration readings in several different ways, the system is subject to component variability (which may be due to extensive component tolerances, degradation, drift, etc.), as described in detail below. Can be difficult. For purposes of this description, assume that each sensor has a variable gain component coupled to its output, and the gain of this component is adjusted to a different setting for each wavelength illuminated by the light emitter. The calibration reading is used to generate a gain setting that varies the output of each sensor so that each wavelength reading corresponds to a predetermined value during calibration.

  Following the calibration operation, the banknote 2 is transported between modules 200A and 200B, and transmission and reflection readings are taken during the evaluation operation. Given N sensors, the banknote is scanned along N longitudinal tracks. Assuming that all color measurements are made at P locations along each track, the total number of reflection measurements made will be C × N × P on each side of the bill. Furthermore, a C × N × P transmission measurement is performed by one sensor of the module. (If necessary, transmission measurements may be performed by sensors of other modules.)

The determination unit 14 uses these measurements to determine the authenticity and face value of the banknote. The measured value is first normalized by the decision unit 14. The measured value is:
Suppose it is represented by Mc _{, n, p, s, t} . Where c represents the wavelength (c = 1, 2,... C), n represents the sensor or track (n = 1, 2,... N), and p represents the position along the track. Representation (p = 1, 2,... P), s represents the bill side (upper or lower) (s = 1, 2), and t represents the type of measurement (reflection or transmission) (t = 1). 2). Examples of normalization algorithms are as follows.

[Example 1]
For each measurement in track n ′ on side s ′, the measurement is normalized using the following function:

  Thus, each measurement is divided by the sum of all other measurements of a group of the same type t ′ (transmission or reflection) on the same side s ′ on the same track n ′ of the banknote, Normalized. (Instead of the sum, the average of these measurements may be used.) Thus, each measurement is normalized with respect to the measurements distributed along (or at least substantially entirely) the same scan line. And in this embodiment means the remaining measurements taken by the same sensor (and using the same group of light emitters).

With this technique, at most C × N × P × S × T normalized measurements M _{NORMc ′, n ′, p ′, s ′, t ′} can be obtained, where S = T = 2 and each normalized measurement represents an individual first measurement M _{c ′, n ′, p ′, s ′, t ′} ).

上式は位置ｎ’ｐ’、ｓ’における色ｃ＝１〜Ｃのタイプｔ’の測定値の平均であり、σ_{ｎ’，ｐ’，ｓ’，ｔ’}は測定値Ｍ_{ｃ’，ｎ’，ｐ’，ｓ’，ｔ’}（ｃ＝１〜Ｃ）の算出された標準偏差（または分散の異なる指標）を表す。 [Example 2]
Alternatively, the measurement may be normalized using the following algorithm:

The above equation is the average of measured values of type t ′ of color c = 1 to C at positions n′p ′ and s ′, and σ _{n ′, p ′, s ′, t ′} are measured values Mc _{′, n. ', P', s ', t'} (c = 1 to C) calculated standard deviation (or an index with different variance).

For example,

  This algorithm has the advantage of normalizing for multiple wavelength measurements at a single location and reducing dispersion in measurements resulting from wear and / or variations in ink density due to different printing conditions.

This technique can also yield up to C × N × P × S × T normalized measurements M _{NORMc ′, n ′, p ′, s ′, t ′} where S = T = 2 and each normalized measurement represents an individual first measurement M _{c ′, n ′, p ′, s ′, t ′} ).

[Example 3]
According to this example, it is divided into groups each containing multiple color measurements of the same type t. The measurements within each group are associated with multiple positions, at least some (and possibly all) of the positions being continuous and may be on the same or opposite sides of the banknote. The group may be different for different target classes. A mapping step is used to derive a set of location identifiers (i = from parameters p, n and s, 1 to 1).

  FIG. 4 shows an example of how a measurement group is derived from parameters p, n and s for one target class. Assuming that the group contains measurements on a single side s = 1, location i = 1 corresponds to the location (s = 1, p = 5, and n = 2) location i = 2 corresponds to the position (s = 1, p = 4, and n = 2, etc.), and the place i = 1 corresponds to the position (s = 1, p = 4, and n = 4). Other groups may be formed as well. Some measurements may be ignored, i.e. not included in any group of the target class. Different groups for individual target face values may be determined empirically in the training process before configuring the evaluation device. Data defining the group may be held in the storage 30.

によって表される。 The measurement is then normalized. Initially, the measurements for each color in each group are averaged. For each group g (g = 1 to G), there are C average measurements, one measurement for each color c ′. Each average measurement is:

Represented by

となる。式中、σ_ｇ’は、グループｇ’の平均化された測定値Ａ_ｃ，ｇ’（ｃ＝１〜Ｃ）の前記算出された標準偏差（または分散の異なる指標）を表す。 The averaged measurements are then normalized in a manner similar to the individual measurements of Example 2. Therefore,

It becomes. In the formula, σ _{g ′} represents the calculated standard deviation (or an index having a different variance) of the averaged measurement values A _{c, g ′} (c = 1 to C) of the group g ′.

For example,

With this technique, it is possible to obtain up to C × G normalized measurements M _{NORMc ′, g ′} (each normalized measurement is an individual step obtained by the spatial averaging method). 1 represents a measured value _{Ac ', g'} ).

によって表される。 [Example 4]
In this example, the techniques of Examples 1 and 2 are combined. Thus, each normalized measurement is:

Represented by

Thus, the measurements are first normalized according to the technique of Example 1 and then further normalized by the technique of Example 2. Thus, this technique can also yield up to C × N × P × S × T normalized measurements M _{NORMc ′, n ′, p ′, s ′, t ′} , where S = T = 2 and each normalized measurement is an amount derived from individual measurements with respect to a measurement group derived from a plurality of different wavelength measurements at positions distributed over a substantial area of the bill By normalizing M _{c ′, n ′, p ′, s ′, t ′} , the individual first measured values M _{c ′, n ′, p ′, s ′, t ′} obtained by themselves. Represents). This procedure may have the advantages resulting from both Examples 1 and 2 above.

As an alternative, the same color measurements from selected locations may be first averaged (similar to Example 3) before normalizing according to the technique of Example 4. With this technique, at most C × G normalized measurements M _{NORMc ′, g ′} can be obtained (each normalized measurement is performed by (a) performing a spatial averaging method. , Deriving a quantity representing a wavelength measurement averaged over a number of positions on the banknote, and then (b) performing a preliminary normalization method to obtain a plurality at positions distributed over a substantial area of the banknote. Obtained from the individual first measurements obtained by normalizing the quantities to a measurement group derived from measurements at different wavelengths).

  In examples 1 and 4, the normalization procedure normalizes the measurements by considering a group of measurements consisting of measurements of the same color and measurements of other colors distributed along the same track. including. Alternatively, the group may include measurements from different areas. The group preferably includes measurements distributed over a substantial area. That is, the group preferably includes measurements from at least 10 locations that are preferably adjacent but may be discontinuous. This area may vary according to the face value of the target. Reservoir 30 may include data defining regions used for normalization in a manner similar to the data defining groups used for averaging in Example 3.

  Note that the use of variance values in Examples 2 to 4 is significantly different from the use of variance values that are part of the data defining acceptance criteria used in known algorithms such as those described in EP-A-0560023. I want to be. The dispersion value is derived from the actual measured value of the banknote under test and represents the range in which the measured property varies with wavelength and / or position. On the other hand, the variance value used in the known algorithm is a stored value representing the range of measured values that vary within the population of individual face value banknotes, not the banknote currently being tested.

  Optionally, multiple sets of normalized measurements using different algorithms, eg, a first set using the algorithm of Example 1 and a second set using the algorithm of Example 2 or 3 It is possible to obtain

式中、Ｍ^＊ _{ｃ’，ｎ’，ｓ’，ｔ’}は、該紙幣の側ｓ’のトラックｎ’におけるタイプｔ’および色ｃ’（正規化された）すべての測定値の（該標的クラスに関する）格納された平均値であり、σ^＊ _{ｃ’，ｎ’，ｓ’，ｔ’}は、これらの（正規化された）測定値に関する対応する格納された分散値である（このような値は両方とも該標的クラスの母集団の測定値から導かれる）。上記操作と同じく、またはそれに代わって、他のデータ圧縮操作を実行してもよい。 After normalizing the measurements, the decision unit 14 uses the normalized measurements along with acceptance criteria for each different target class defined by the data stored in the storage 30 so that the bill is the target Determine if you belong to one of the classes. This can be achieved using a variety of different techniques known per se. However, the selected technique preferably involves at least in part determining whether the relationship between the various measurements matches a known correlation, for example as determined by a training operation. For example, for each target class, features that are processed using data from the storage 30 representing the inverse covariance matrix and the mean value associated with that target class to combine the measurements to derive the Mahalanobis distance. A vector may be formed. If the Mahalanobis distance is less than a predetermined value, the bill is considered to belong to that target class. Alternatively, the measurements are processed using data for another target class. A data compression operation is preferably performed when deriving the feature vector to reduce the dimensionality of the vector. For example, the measurements associated with each color in each scanned line may be combined, thereby reducing the number of dimensions by a factor P. One way to accomplish this is to take the absolute value of the difference between the measured value and the average value stored in the storage 30 and divide by the variance value stored in the storage 30 and then average the result Would take. Thus, using the measured value, a vector having each of its C × N × 2 × 2 dimensions represented by the following equation is derived:

Where M ^* _{c ′, n ′, s ′, t ′} is the type t ′ and color c ′ (normalized) of all measurements in track n ′ on the side s ′ of the banknote (the target Σ ^* _{c ′, n ′, s ′, t ′} are the corresponding stored variance values for these (normalized) measurements (such as Both values are derived from measurements of the target class population). Other data compression operations may be performed in the same manner as or in place of the above operations.

  To reduce the number of calculations, a measurement (or a value derived by combining measurements) can be compared to an upper or lower threshold associated with the target class, and if that measurement falls within that threshold As long as the Mahalanobis distance calculation can be performed.

  As another example, normalized measurements can be processed and obtained using a set of individual coefficients derived by training a neural network with a sample of a target class. Value can be examined to determine if the bill belongs to that class.

  Thanks to the technique used in the above embodiment, the normalized measurements used to evaluate the banknote show little variation as a result of component differences, and the variance due to banknote degradation and dirt is also low. Almost no. As a result, acceptance criteria allow for enhanced recognition and identification. This is accomplished while allowing calibration in a transmission system using a simple structure.

  Various modifications are possible. For example, in the above embodiment, calibration measurements were used to control the gain applied to the sensor output during the evaluation operation. Alternatively, using calibration measurements, (a) determine the intensity with which the light emitter illuminates the banknote during the evaluation operation, or (b) determine the amount of digital adjustment of the measurement obtained during the evaluation operation. Can be determined.

  Two or more of these possibilities may be combined. For example, calibration measurements can be used to set the emitter intensity as high as possible without saturating the sensor that receives the emitted light. In addition, further calibration measurements made at this setting are used to derive a coefficient (representing the ratio of the predetermined value to the actual calibration measurement), which is then used during the evaluation (before its normalization). The associated measurement taken may be corrected.

It is a block diagram which shows the automatic transaction machine incorporating the banknote evaluation apparatus by this invention. 1 is a schematic diagram showing a part of a measurement unit of the evaluation apparatus. It is a schematic sectional drawing which shows the optical unit of this measurement unit. It is a figure which shows how the measured value of the different area | region of a banknote is grouped for normalization.

Claims (16)

A method for inspecting banknotes, the method comprising:
Measuring at each different position of the banknote to determine the optical properties of the banknote at different wavelengths;
Normalizing the measurement;
Applying an acceptance criterion to the normalized measurement to determine whether the normalized measurement represents a target banknote class, and performing an evaluation operation with
Each measurement is normalized to a measurement group consisting of a plurality of sets of measurements each including measurements at different wavelengths at locations distributed over a substantial area of the banknote ;
Performing a calibration operation to obtain a calibration measurement that is independent of the banknote and influencing the evaluation operation according to the calibration measurement to compensate for component variations;
The calibration operation comprises:
Operating a light emitter on one side of the bill passage;
Detecting light emitted by the light emitter and reflected from a window on the opposite side of the bill passage using a sensor disposed on the one side of the bill passage . Measurement of each set are distributed along the least even entire length of the scan lines of the bill, method of claim 1. Each measurement group may be formed by a plurality of sets of measurements derived entirely by scanning a common line, the method according to claim 2. Measurements of each group, the measurement of the other group are performed using different individual sensors A method according to any one of claims 1 to 3. A method for inspecting banknotes, the method comprising:
Measuring and determining the optical properties of the banknote at different wavelengths at each different position of the banknote;
Deriving a normalized measurement value from the measurement and applying acceptance criteria to the normalized measurement value to determine whether the normalized measurement value represents a target banknote class; Performing an accompanying evaluation operation,
The normalized measurement is
(I) classifying the first measurement values obtained from the measurement into measurement groups each including measurement values relating to different wavelengths at at least one position of the banknote;
(Ii) deriving normalized measurement values each representing a relationship between an individual first measurement value and a variance value representing the variance of the first measurement value in the group to which the first measurement value belongs; By
Led and
Performing a calibration operation to obtain a calibration measurement that is independent of the banknote and influencing the evaluation operation according to the calibration measurement to compensate for component variations;
The calibration operation comprises:
Operating a light emitter on one side of the bill passage;
Detecting light emitted by the light emitter and reflected from a window on the opposite side of the bill passage using a sensor disposed on the one side of the bill passage . Each normalized measurements, (i) representing the difference and (ii) the ratio between the variance value of the first measurement value corresponding to the average of the measured values of the measurement group, as claimed in claim 5 Method. Each first measurement is derived by averaging the measurements of the individual wavelengths in has been selected positions distributed on the bill, method of any one of claims 5 or 6. Each first measurement is a measurement group derived from a plurality of different wavelength measurements at positions distributed over a substantial area of the banknote, an amount representing a wavelength measurement at at least one position on the banknote. guided by the normalization operation for normalized to a method according to any one of claims 5, 6 or 7. The measurement performed in the evaluation operation is a reflection measurement performed using the light emitter and the sensor; (i) an optical device under the window; and (ii) one of the light emitter and sensor. and a transmission measurement using light traveling between a method according to any one of claims 1 to 8. A method for inspecting banknotes, the method comprising:
Reflection measurements are made using a light emitter and sensor placed together on one side of the banknote path and transmitted using an optical device on the other side of the banknote path along with either the light emitter or the sensor. Measuring, thereby determining the optical properties of the banknote at each different position of the banknote;
Applying an acceptance criterion to the measurement to determine whether the measurement represents a target banknote class, and performing an evaluation operation with
The method further includes performing a calibration operation to obtain a calibration measurement independent of the banknote and influencing the evaluation operation in accordance with the calibration measurement to compensate for component variations;
The calibration operation comprises:
Operating the light emitter;
Said emitted from the light emitting device, located on the other side of the paper passageway, the light reflected from the window that is present on said optical device, and a step of detecting using said sensor, method. The bill is disposed along a line transverse to the direction to be scanned, including a plurality of optical devices having a common window, the method according to any one of claims 1 to 10. Having a flat surface on which the window is facing the emitter and sensor, the method according to any one of claims 1 to 11. The step of affecting the evaluation operation comprises:
(A) controlling the intensity with which the light emitter illuminates the banknote during an evaluation operation according to the calibration measurement;
(B) controlling the gain applied to the output of the sensor during an evaluation operation according to the calibration measurement;
(C) the calibration comprises a measurement evaluation measurements acquired during operation according to one or more of the digitally adjustable to process, the method according to any one of claims 1 to 12. A plurality of light emitters and a plurality arranged in an evaluation operation so as to allow a plurality of measurements to be made from banknote properties of different wavelengths at positions distributed along a line traversing the banknote scanning direction. when used in the evaluation device comprising a sensor, the calibration operation, the light emitter and with a sensor, comprising the step of performing a calibration measurement corresponding to the position and wavelength, respectively measured in the evaluation operation, claim the method according to any one of 1 to 13. The evaluation operation includes performing reflection measurements on each side of the banknote path using individual light emitter / sensor pairs, such one pair of light emitters and such other pair of sensors also being transmitted. used to perform the measurement method according to any one of claims 1 to 14. 16. A device for inspecting banknotes, the device configured to operate according to the method of any one of claims 1-15 .

Images